Lanparty NFII Ultra B raising FSB

[TheGame240]

I've had this board for a couple of months now, but I've been held back by my crappy Geil value ram. Worked fine at 200mhz (what it's rated for), but wouldn't go a bit past that. With the urge to see what this board was capable of, I picked up some used BH5 memory. Well, it arrived today and I immediately began to see what it could do.

 

I started off by trying to see what the ram could do by itsself. I was a tad disappointed, but not completely. The previous owner said he had it running 240 with 2.9v at 11-2-2-2 timings so that's where I started. Memtest errored at test 4. So I upped the voltage to 3.2 and passed a complete loop of memtest at that speed. Anything past 240 would give errors in memtest so I upped the voltage to 3.3v and managed 250 error free in memtest.

 

I then decided to boot into Windows and see what I could actually get out of the board itsself. Started out with 240 (figured I'd go big) with a multiplier of 7 and 1.6v on the CPU and 3.3 vdimm. Well, 240 boots into Windows, but reboots trying to run Prime95. At that point the chipset voltage was 1.7 and knowing that the memory was stable enough to run 250 I figured the chipset was holding me back so I upped the voltage to 1.8. Well this time Prime95 failed at test 1, but no reboot. So I upped the voltage again to 1.9 and this time Prime95 failed, but not at fast. So I'm guessing I'm being held back by the MB at this point. Right now I'm running Prime95 at 1.6v on the chipset just so I can say I've tried everything voltage wise there. *EDIT* Well 1.5 failed, but it took 25 minutes which is longer than any of the other voltages ran (go figure :confused: ).

 

Temps are ok, because for testing purposes I placed a house fan blowing into the case, so I could rule out overheating as the cause of any of my problems. The normally very warm south bridge is actually pretty cool to the touch. Ram is currently 33C and the processor is running 30C socket and 20C diode ( I know they're not accurate, but they're not 30 degrees off either ).

 

Anyway, I'm at my limit as to what else to try. I'm getting a rounding error in Prime95 if that helps anyone. I think I'm going to try setting everything back at 250 and running memetest overnight just to make sure the ram is 250 stable. If anyone has any suggestions for getting the FSB up some more (250 would be the sweet spot), or just stabalizing it at 240, I'd love to hear them.

[SAE]

Try to change slots... when using 1&3 try 2&3... or the other way around. Try to change slots for the single sticks, in my experience each stick has its favourite slot.

 

Try to use other alpha settings... I used 2-0-3-3-1-0-3 for the latest NF2 benches... (before I sold it).

 

A good burn in helps a lot too.

You said, the person you got the ram from was running it 240 at 2.9V. I guess he had it running in an A64 board. Those tend to clock ram much better due to shorter signal way to the integrated mem controller (very low latencies).

 

Try to let the ram burn in for 24hours at 3.3V and 250MHz fastest timings possible... (11-2-2-2-2.0-9-12-e-e-f-e-2-0-3-3-1-0-3)

 

That may help as well. Furthermore several board/ram combinations seemed to like lower vdd voltages, some also needing 2V+ for vdd.

 

G'luck mate.

[TheGame240]

I'll give those timings a try and if that doesn't work I'll try playing around with different slots and see what kind of results I get there.

 

The think I find pretty odd is that Prime95 lasted long with 1.6v on the chipset than with any other voltage, which leads me to believe that maybe it isn't being cooled properly. Give the house fan I have blowing directly into the case, I'm guessing that maybe the HS isn't attached properly? Does anyone have any first hand knowledge about how well DFI attaches the sinks to these boards?

 

Also I left Memtest running overnight at 250 with 0 errors.

[red930]

Both nb + sb on my board had good contact w/thermal paste.Put NB-1 on and put old dfi nb sink on sb w/40mm fan using ceramique ,way cooler sb temps ,more stable

[TheGame240]

Both nb + sb on my board had good contact w/thermal paste.Put NB-1 on and put old dfi nb sink on sb w/40mm fan using ceramique ,way cooler sb temps ,more stable

 

That's good to know, since I really didn't want to go through pulling the motherboard just to check the sinks.

 

Just to give an update, I can boot into Windows at 246, but it's not stable. It stable enough for some basic things (I'm typing this at 246), but I can't pass Prime at all. The highest I've had it Prime stable was 230.

 

Oddly enough the only way to get it to pass Prime at 230 was at 3.3vdimm. So that's leading me to a couple of conclusions.

 

1. The heat spreaders on the ram are acting more like heat trappers and even with the house fan too much heat is building up. I think heat might be an issue because at 240 if I turn the fan off I'll get errors in memtest at 3.3v and the temp (probe is attached to the spreaders) will get up to 49C. Is that a pretty average temp for ram with that voltage at that speed with no fan? If it seems low, then that would further support my theory that the chips themselves are actually much hotter and the spreader isn't doing what it should be.

 

That leads to a couple of more questions. Is it ok to remove the spreaders and just run a fan over the ram? I'm considering this bracket so I can position the fan over the ram. Or are ramsinks/heatspreader a must have with BH5 at higher voltages.

 

2. My second theory is that somewhere along the way my 3.3v rail is dropping. I've measured the rails with a DMM at the clips under a 3dMark05 loop. The 12v line measures 11.98, 3.3v measures dead on, and the 5v line measures 4.98. But then I started thinking that there's probably a lot of circuitry that everything has to go through before the ram actually gets it's voltage. So maybe with everything up and running the ram is getting undervolted. This would explain a couple of oddities.

a. I'm able to loop memtest at 250 with no errors, but 230 is the highest I can Prime.

b. The previous owner had the ram stable at 240 with 2.9v on an NF7 with the same timings I do, but I need at least 3.2v to pass Memtest loops at 240.

c. I tried setting the voltage to 2.8v (default max) and the FSB to 200 and it wasn't Prime stable.

 

That's where I'm at now. This weekend I'll play around with trying single sticks and different slots, but for now I'd like to get some more input on where the likely problem is. If anyone knows where on the MB I could measure the actual vdimm that would be great too (and pictures always help). If it turns out to just be a low 3.3v rail then I'll need a new power supply and I can stop further testing till then, so I want rule that out first.

 

Also some input on whether or not I should risk removing the head spreaders would be appreciated as well.

[TheWizzard]

Okay, okay, listen :nod:

 

Let me, from my own experience, give you some tips.

 

1. Maybe it's a good idea to sink the board; cpu mosfets, dimm mosfets, agp mosfets. Somewhere in this forums is a picture with all the important hotspots to sink and to cool.

 

2. Once sinked, place a fan on the mosfet area and a fan on the rams (so 2 separate fans for each area).

 

3. Maybe it's a better idea to replace the standard sink of the nb with an actively cooled one.

 

4. Replace the standard sb sink with the standard nb sink.

 

5. Attaching all the sinks on the board can be done by thermal tape or by this method (wich is imho more efficient): place a tiny drop of superglue on the edges of the chips to be sinked, put a little drop of as5 or as ceramique in the middle of the chip and attach the sink that way. Then place a fan on that spot.

 

6. What bios are you using. On my board the latest hellfire turbo bios works very well, maybe it does the same on yours.

 

7. What volts are you using for the cpu? Prime unstability can also occur due to too less cpu juice.

 

8. As SAE already told. Try moving the rams in the slots (but keep it in dual channel). I personally recommend slot1 & 3. This works best for me. We have the same ram.

 

This is pretty much what comes to my mind for now. If anyone can add something to this list, please feel free to add the numbers.

[TheGame240]

Okay, okay, listen :nod:

 

Let me, from my own experience, give you some tips.

 

1. Maybe it's a good idea to sink the board; cpu mosfets, dimm mosfets, agp mosfets. Somewhere in this forums is a picture with all the important hotspots to sink and to cool.

 

I'll search for that and see what I can come up with. I think I have some old K6 HSs I can cut up for that.

 

2. Once sinked, place a fan on the mosfet area and a fan on the rams (so 2 separate fans for each area).

 

Ok. That shouldn't be too much of a problem. So two brackets then?

 

3. Maybe it's a better idea to replace the standard sink of the nb with an actively cooled one.

 

I'll probably hold off on that one until I'm sure it needs to be tried. In your experience, how well was the stock cooling? Do you think adding a fan to the NB would be sufficient?

 

4. Replace the standard sb sink with the standard nb sink.

 

If/when I get to 3, this will definately be a step.

 

6. What bios are you using. On my board the latest hellfire turbo bios works very well, maybe it does the same on yours.

 

6/19 official

 

7. What volts are you using for the cpu? Prime unstability can also occur due to too less cpu juice.

 

1.625 for the CPU, which is stable at 2.2 with 1.6v. I used a multi of 7 for all FSB testing to ensure the CPU wasn't a factor.

 

8. As SAE already told. Try moving the rams in the slots (but keep it in dual channel). I personally recommend slot1 & 3. This works best for me. We have the same ram.

 

That will be what I'll be trying this weekend. In your experience with this ram, are the stock heat spreaders adequate for cooling, or should I look for ram sinks as well?

[red930]

ExtremeOC has psu testing guide under power supplies /testing your psu,pretty informative I wouldn't take spreaders off ram till last resort ,get fan over ram 1st(I have 2 80mm fans wired together rubberbanded to ram slot retaining levers,temporary testing utt memory at 3.4v) I have to run 1.85 - 1.875v vcor to get to 250x10 with tccd mem 2.5,3,3,11 and 244x10 w/twinmos utt 2,2,2,8 (about to try 10.5)

[TheGame240]

Update. After playing around all weekend with various configurations and some modded bios, I've come to the conclusion either:

 

1. The power supply is suspect.

 

2. Heat is the issue due to improperly mounted HS.

 

3. The board is a dud.

 

So far the highest STABLE FSB I've been able to reach is 210. I have, however, been able to boot into windows and do some basic things as high as 246.

 

I suspect maybe I'm hitting a wall with the power supply and tripping overcurrent protect since lots of settings under stress will cause a reboot. Measureing voltages under load with a DMM shows the 12v rail going from 11.88-11.85, 5v rail going from 5-4.96 and 3.3v going from 3.32-3.29 so voltage drops aren't the problem, but I suspect I'm either drawing too much current on the 3.3 and 5v rails or on the 12v rail. I'm thinking probably the 12v rail since this seems to show the lowest readings and is rated for the least amount of current. PSU specs can be found here.

 

I also suspect NB heat build up because the HS itsself (top of it) is much cooler than the surrounding board and area around the base. So obviously heat isn't being transfered like it should. I've ruled out the ram overheating since when I slide the temp probe between the spreader and board next to the chips the temps are pretty much the same as on the outside of the spreader so to me that indicates the spreader isn't trapping heat.

 

So my next step is to:

 

1. Replace PSU

 

2. Replace NB HS (moving it to the SB is not an option though because of the 6800)

 

3. Add a fan over the ram using the bracket as stated above.

 

Now I need some suggestions on new power supplies and replacement coolers. I definately want to try to be as futureproof as possible with the new PSU, so I'm thinking 500w is a must have. What about dual 12v rails. I've read that's a benefit for newer boards (especially SLI boards), but is it a hinderance to NF2 boards?

 

So far the only ones that have really caught my eye were

http://www.newegg.com/app/ViewProductDesc....-104-982&depa=0

http://www.newegg.com/app/ViewProductDesc....-104-934&depa=0

http://www.newegg.com/app/ViewProductDesc....-103-478&depa=0

http://www.newegg.com/app/ViewProductDesc....-104-152&depa=0

 

I've read that Fortron has a great $$/performance ratio. I do like the 5 year warranty of the OCZ along with the adjustable rails though. Does anyone know if you can get blue leds on the OCZ? Anyway the OCZ PS is probably at the upper end of my budget, so if I won't really know the difference I'll probably go with the Fortron or Sparkle PSU.

 

For the chipset does anyone know of anything better than this one?

http://www.newegg.com/app/ViewProductDesc....-109-116&depa=0

 

I would just like to thank everyone for their advice so far an hopefully I can get some positive input on a new PSU and chipset cooler and start torturing this board with a little more success.

[red930]

As you can see us 32biters don't draw much attention in the ocing forum and I guess I can understand that.I guess I'll try to shed some light w/my limited understanding.

#1 I think your on the righttrack here,just looking at your psu specs it looks like you get 120w to the 12v rail,your 5 drives could eat up 100w but they'r not all going at once ,I would think your cpu 80w at least if your ocing(think i remember looking in sandra and seeing over 100w being used and I'm not sure what rail the fans use

#2 heat issue if your talking about the nb,heat is seldom a factor here,all the same I've a nb-1 on mine just so I could put old nb on sb w/fans on both.

#3 board is dud I doubt it,not enough test yet to confirm,most of the time the user wants to blame the board when in reality the user just hasn't armed himself w/enough knowledge

These boards are thoroughbreds built for ocing the more you learn about them the more you get out.The best part is they're older and all the testing has already been done and documented,we just have to look for it (forum search).The guys on newer stuff got it harder they're still trailblazin

It looks like you got some nice memory there,problem is it might not be that compatible running dc in 2x512 seems like I remember reading 2x256 bh5 was great but 2x512 bh5 not so great,but maybe thats a myth(thats for this chipset)bh5 still highly desirable

Lookin at your dmm measurements they are still within spec,but is that idle or load

For the power supply around here the Fortrons are highly regarded for nf2 but to be future proof you might want to read thishttp://forums.extremeoverclocking.com/showthread.php?t=136602 and just to keep yourself busy arming yourslf w/knowledge http://forums.pcper.com/showthread.php?p=2592157&postcount=9 and this http://www.ocforums.com/showthread.php?t=325031

[red930]

Sorry about that ExtremeOC link but it is a pretty in depth look at a topic your interested in its in the psu section(EOC in general is a good forum to search in) I'd be more worried about the sb it is more notorious for overheating than nb and use nonductive TIM like AS arctic alumina thermal adhesive,I use a 50/50 mix of that and Ceramique but whatever either will work

[Poisonsnak]

1.625 for the CPU, which is stable at 2.2 with 1.6v. I used a multi of 7 for all FSB testing to ensure the CPU wasn't a factor.

 

I don't really agree with this, I realize you are running a super low internal clock but the CPU voltage requirements go up with both FSB and internal clock, this means that 250x10 requires more voltage than 200x12.5. You might say "well even though I'm pushing 240fsb my CPU is at 1700 or so, so my volts can be allocated to fsb" but I don't think that's how it works. If you're memtest stable at a given FSB and then prime dies on you I would most definitely point at the CPU.

 

I would personally slam the CPU with something like 1.80 this is relatively safe even on medium-level air cooling.

 

On the PSU front I hit my 250x10 with the Antec True380S (18A@12V) so I don't know if I would point at the PSU, although you have a lot of drives and a pretty huge video card so who knows.